[Bug tree-optimization/33434] [4.3 Regression] inlining miscompilation

[hubicka at gcc dot gnu dot org]

------- Comment #15 from hubicka at gcc dot gnu dot org  2007-11-08 12:42 -------
The problem here is that code assumes that variable without DEFAULT_DEF is not
in SSA.  Arguments that are unused also do have DEFAULT_DEF=NULL, this patch
should fix it. I am testing it now.

Honza
Index: tree-inline.c
===================================================================
*** tree-inline.c       (revision 129992)
--- tree-inline.c       (working copy)
*************** setup_one_parameter (copy_body_data *id,
*** 1420,1425 ****
--- 1420,1434 ----
        && !useless_type_conversion_p (TREE_TYPE (p), TREE_TYPE (value)))
      rhs = fold_build1 (NOP_EXPR, TREE_TYPE (p), value);

+   /* If the value of argument is never used, don't care about initializing
it.
+      This is needed for correctness since we otherwise confuse it with
non-SSA
+      variable later and end up renaming otherwise.  */
+   if (gimple_in_ssa_p (cfun) && is_gimple_reg (p) && !def)
+     {
+       gcc_assert (!TREE_SIDE_EFFECTS (value));
+       return;
+     }
+ 
    /* If the parameter is never assigned to, has no SSA_NAMEs created,
       we may not need to create a new variable here at all.  Instead, we may
       be able to just use the argument value.  */


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33434